Abstract: To address the stability degradation of permanent magnet synchronous motors (PMSM) caused by uncertain load disturbances during operation, a novel sliding mode observer (SMO) is developed. This observer integrates the strengths of linear and nonlinear super-twisting algorithms and leverages an optimized sparrow search algorithm (SSA) for parameter tuning. The optimization algorithm employs cubic chaotic map and oppositional learning for population initialization, mitigating the issue of homogeneous initial positions. Additionally, Gaussian mutation is applied to adjust follower positions, enhancing the algorithm’s capability to escape local optima. During motor startup or sudden load changes, threshold-based monitoring of current or torque variations is introduced, dynamically adjusting the weights of the fitness function to reduce oscillations or overshoot, thus improving the system’s responsiveness to torque changes. Compared with traditional super-twisting sliding mode observers, simulation results indicate that the optimized observer reduces the estimation error of load torque by 44% and decreases speed fluctuations during load disturbances by 9.26%.